Slack AI

Slack AI brings Bayes, Monte Carlo's AI teammate, into Slack through the Bayes app, so your team can investigate data and AI reliability issues without leaving the conversation. Mention @Bayes in a channel, reply in an alert thread, or send the app a direct message to ask Bayes questions in plain language, triage and troubleshoot alerts, check the health of your tables, and even build monitors.

Slack AI is a superset of the classic Slack notifications integration. It delivers Monte Carlo alerts to your channels and lets you act on them with built-in message buttons — triage, acknowledge, resolve — exactly like the classic app, and it adds Bayes on top, so you can talk to Monte Carlo in natural language and have Bayes investigate on your behalf. You don't need both: installing Slack AI on its own gives you everything the classic Slack app does, plus conversations with Bayes. Alert delivery through Slack AI must be enabled for your account. See Deliver alerts.

What you can do

Mention @Bayes in a channel or send the app a direct message to have Bayes:

• Ask about your data — "How's the health status of table X?", "When was table Y last updated?", "Are there any alerts assigned to me today?"
• Troubleshoot alerts — ask Bayes to investigate an alert and get a summary of likely root causes, posted right in the thread.
• Triage incidents — have alerts assessed and prioritized so your team knows what to look at first.
• Build monitors — describe what you want to watch for and Bayes proposes the monitor logic and SQL.
• Take action — add a comment to an alert, check what a data issue would impact, and more.

Slack AI connects you to the same Bayes you use in the Monte Carlo UI. Bayes answers directly or uses the matching skill for your request, such as troubleshooting, triage, cost & performance, or PII detection.

Requirements

Using Bayes in Slack's AI agent panel (the app container) requires a paid Slack plan. @mentions, direct messages, and alert notifications work on any Slack plan. See Slack's requirements for AI agent apps.

You will also need permission to add integrations in Monte Carlo (or an administrator who can complete the install), and permission to install apps in your Slack workspace — you may need your Slack administrator to approve the request.

Set up Slack AI

Installing Slack AI requires a Monte Carlo account, since the install starts from Monte Carlo's Settings page and redirects to Slack from there.

1. Go to Settings > Integrations.
2. Click the Add button
3. In the Collaboration section, click Slack AI.

1. You will be redirected to Slack to install and authorize the Bayes app. Review the requested permissions and approve them. You may need to wait for your Slack administrator to approve the request.
2. After you authorize the app, Slack confirms the installation and you are returned to Monte Carlo, where the new Slack AI connection appears in your integrations list.

Once connected, the app's Home tab in Slack confirms your account is linked and shows how to start.

Each Slack user can be linked to one Monte Carlo account installation at a time. Use Disconnect on the Home tab to unlink before connecting a different account.

Start using it in Slack

Once the app is installed:

1. In Slack, open a direct message with the Bayes app, or invite it to a channel with /invite @Bayes.
2. Ask a question or describe what you need — for example, "Troubleshoot the freshness alert on analytics.orders." You can also reply directly in an alert thread.
3. Bayes replies in the thread. For deeper investigations it includes a link to open the full findings in Monte Carlo.

Workspace settings

Each connected Slack workspace has its own settings. Open Settings > Integrations, click the Slack AI connection, and find them under Integration details. Changing them requires permission to edit integrations.

Deliver alerts

Controls which Slack app delivers this workspace's Monte Carlo notifications.

Slack AI delivers alerts on its own. It does not need the classic Slack app. Alerts arrive with the same Triage, Acknowledge, and Resolve buttons, plus Bayes in the thread. To post in a public channel you chose as a notification target, the app joins that channel.

Alert delivery through Slack AI must be enabled for your Monte Carlo account. Contact Support to enable it.

When Slack AI is the only Monte Carlo app in the workspace, it delivers alerts once the feature is enabled, and there is nothing to configure.

When the workspace also has the classic Slack app, this setting picks which app delivers. Turn it on to have Slack AI deliver the notifications for every audience pointed at this workspace, in place of the classic app.

Turning it on is a request rather than an immediate switch. Delivery also depends on the workspace's connection health and on Slack having granted the app permission to join channels. Until those are satisfied, alerts keep arriving through the classic app and Monte Carlo shows a warning next to the setting. Contact Support if the warning persists.

Sample data

Controls whether Bayes may read sample rows from your warehouse while answering in this workspace.

This setting is on by default. Sample rows help Bayes explain a data quality issue concretely — for example, showing which values broke a validation rule instead of only reporting that one did.

Turn it off if you would rather no raw row values reach Slack from this workspace. Bayes then answers from metadata, monitor results, and aggregates alone, and the restriction applies for the whole conversation, including any skill or tool Bayes uses for the request.

Two things this setting does not change:

• It does not stop Monte Carlo from collecting samples. What Monte Carlo samples is governed by your warehouse and table configuration, independently of Slack. This setting governs whether Bayes may read those samples while answering in Slack.
• It does not widen anyone's access. It can only restrict. A user who cannot see sample data in Monte Carlo still cannot see it through Slack AI, whatever this setting says.

The setting is per workspace, so you can connect a workspace for a team that should see sample values and another that should not.

Permissions and data

In Slack, Bayes runs with the same permissions as the person interacting with it — it cannot access data or take actions that the user could not perform in Monte Carlo directly.

Raw sample rows are covered by an additional workspace-level control; see Sample data.

Slack scopes requested

When you install the app, Slack asks you to approve the bot token scopes below. Monte Carlo only reads content in conversations the app has been added to.

Slack admins can decline the two optional scopes, files:read and groups:history. The app still works without them:

• Without files:read, Bayes cannot read files shared in Slack.
• Without groups:history, Bayes cannot read messages in private channels. It still answers when you @mention it there.

The app also subscribes to these Slack events: app_home_opened, app_mention, app_uninstalled, message.channels, message.groups, message.im, reaction_added, reaction_removed, and tokens_revoked.

Workspaces that installed the app before these scopes were added must reinstall it to approve them. Start the install again from Settings > Integrations.

For details on what data Monte Carlo collects and how it is used, retained, and deleted — and how to make access or deletion requests — see the Monte Carlo Privacy Policy.

For details on how Monte Carlo secures and governs its AI features — including data handling and the models used — see AI Security & Governance.

Support

Need help? Reach the Monte Carlo team by asking Bayes.

Remove Slack AI

To disconnect Slack AI, go to Settings > Integrations, open the Slack AI connection, and remove it. You can also uninstall the Bayes app from your Slack workspace's app management settings.

Common Use Cases